From d63c277dc672e0c568481af043359420fa9d4736 Mon Sep 17 00:00:00 2001 From: Mario Kleiner Date: Wed, 29 Mar 2017 22:09:11 +0200 Subject: drm/amdgpu: Make display watermark calculations more accurate Avoid big roundoff errors in scanline/hactive durations for high pixel clocks, especially for >= 500 Mhz, and thereby program more accurate display fifo watermarks. Implemented here for DCE 6,8,10,11. Successfully tested on DCE 10 with AMD R9 380 Tonga. Reviewed-by: Alex Deucher Signed-off-by: Mario Kleiner Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdgpu/dce_v6_0.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/dce_v6_0.c') diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c index 838cf1a778f2..b20cc07eea64 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c @@ -986,7 +986,7 @@ static void dce_v6_0_program_watermarks(struct amdgpu_device *adev, struct drm_display_mode *mode = &amdgpu_crtc->base.mode; struct dce6_wm_params wm_low, wm_high; u32 dram_channels; - u32 pixel_period; + u32 active_time; u32 line_time = 0; u32 latency_watermark_a = 0, latency_watermark_b = 0; u32 priority_a_mark = 0, priority_b_mark = 0; @@ -996,8 +996,8 @@ static void dce_v6_0_program_watermarks(struct amdgpu_device *adev, fixed20_12 a, b, c; if (amdgpu_crtc->base.enabled && num_heads && mode) { - pixel_period = 1000000 / (u32)mode->clock; - line_time = min((u32)mode->crtc_htotal * pixel_period, (u32)65535); + active_time = 1000000UL * (u32)mode->crtc_hdisplay / (u32)mode->clock; + line_time = min((u32) (1000000UL * (u32)mode->crtc_htotal / (u32)mode->clock), (u32)65535); priority_a_cnt = 0; priority_b_cnt = 0; @@ -1016,7 +1016,7 @@ static void dce_v6_0_program_watermarks(struct amdgpu_device *adev, wm_high.disp_clk = mode->clock; wm_high.src_width = mode->crtc_hdisplay; - wm_high.active_time = mode->crtc_hdisplay * pixel_period; + wm_high.active_time = active_time; wm_high.blank_time = line_time - wm_high.active_time; wm_high.interlaced = false; if (mode->flags & DRM_MODE_FLAG_INTERLACE) @@ -1043,7 +1043,7 @@ static void dce_v6_0_program_watermarks(struct amdgpu_device *adev, wm_low.disp_clk = mode->clock; wm_low.src_width = mode->crtc_hdisplay; - wm_low.active_time = mode->crtc_hdisplay * pixel_period; + wm_low.active_time = active_time; wm_low.blank_time = line_time - wm_low.active_time; wm_low.interlaced = false; if (mode->flags & DRM_MODE_FLAG_INTERLACE) -- cgit v1.2.3 From e190ed1ea7458e446230de4113cc5d53b8dc4ec8 Mon Sep 17 00:00:00 2001 From: Mario Kleiner Date: Wed, 29 Mar 2017 22:09:12 +0200 Subject: drm/amdgpu: Avoid overflows/divide-by-zero in latency_watermark calculations. At dot clocks > approx. 250 Mhz, some of these calcs will overflow and cause miscalculation of latency watermarks, and for some overflows also divide-by-zero driver crash ("divide error: 0000 [#1] PREEMPT SMP" in "dce_v10_0_latency_watermark+0x12d/0x190"). This zero-divide happened, e.g., on AMD Tonga Pro under DCE-10, on a Displayport panel when trying to set a video mode of 2560x1440 at 165 Hz vrefresh with a dot clock of 635.540 Mhz. Refine calculations to avoid the overflows. Tested for DCE-10 with R9 380 Tonga + ASUS ROG PG279 panel. Reviewed-by: Alex Deucher Signed-off-by: Mario Kleiner Signed-off-by: Alex Deucher Cc: stable@vger.kernel.org --- drivers/gpu/drm/amd/amdgpu/dce_v10_0.c | 19 +++---------------- drivers/gpu/drm/amd/amdgpu/dce_v11_0.c | 19 +++---------------- drivers/gpu/drm/amd/amdgpu/dce_v6_0.c | 19 +++---------------- drivers/gpu/drm/amd/amdgpu/dce_v8_0.c | 19 +++---------------- 4 files changed, 12 insertions(+), 64 deletions(-) (limited to 'drivers/gpu/drm/amd/amdgpu/dce_v6_0.c') diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c index 67e7d8ee1aab..2b78d58c9c35 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c @@ -1090,23 +1090,10 @@ static u32 dce_v10_0_latency_watermark(struct dce10_wm_params *wm) a.full = dfixed_const(available_bandwidth); b.full = dfixed_const(wm->num_heads); a.full = dfixed_div(a, b); + tmp = div_u64((u64) dmif_size * (u64) wm->disp_clk, mc_latency + 512); + tmp = min(dfixed_trunc(a), tmp); - b.full = dfixed_const(mc_latency + 512); - c.full = dfixed_const(wm->disp_clk); - b.full = dfixed_div(b, c); - - c.full = dfixed_const(dmif_size); - b.full = dfixed_div(c, b); - - tmp = min(dfixed_trunc(a), dfixed_trunc(b)); - - b.full = dfixed_const(1000); - c.full = dfixed_const(wm->disp_clk); - b.full = dfixed_div(c, b); - c.full = dfixed_const(wm->bytes_per_pixel); - b.full = dfixed_mul(b, c); - - lb_fill_bw = min(tmp, dfixed_trunc(b)); + lb_fill_bw = min(tmp, wm->disp_clk * wm->bytes_per_pixel / 1000); a.full = dfixed_const(max_src_lines_per_dst_line * wm->src_width * wm->bytes_per_pixel); b.full = dfixed_const(1000); diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c index 7a91afba0198..fec2a69f3ae2 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c @@ -1059,23 +1059,10 @@ static u32 dce_v11_0_latency_watermark(struct dce10_wm_params *wm) a.full = dfixed_const(available_bandwidth); b.full = dfixed_const(wm->num_heads); a.full = dfixed_div(a, b); + tmp = div_u64((u64) dmif_size * (u64) wm->disp_clk, mc_latency + 512); + tmp = min(dfixed_trunc(a), tmp); - b.full = dfixed_const(mc_latency + 512); - c.full = dfixed_const(wm->disp_clk); - b.full = dfixed_div(b, c); - - c.full = dfixed_const(dmif_size); - b.full = dfixed_div(c, b); - - tmp = min(dfixed_trunc(a), dfixed_trunc(b)); - - b.full = dfixed_const(1000); - c.full = dfixed_const(wm->disp_clk); - b.full = dfixed_div(c, b); - c.full = dfixed_const(wm->bytes_per_pixel); - b.full = dfixed_mul(b, c); - - lb_fill_bw = min(tmp, dfixed_trunc(b)); + lb_fill_bw = min(tmp, wm->disp_clk * wm->bytes_per_pixel / 1000); a.full = dfixed_const(max_src_lines_per_dst_line * wm->src_width * wm->bytes_per_pixel); b.full = dfixed_const(1000); diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c index b20cc07eea64..6604bcf783b7 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c @@ -861,23 +861,10 @@ static u32 dce_v6_0_latency_watermark(struct dce6_wm_params *wm) a.full = dfixed_const(available_bandwidth); b.full = dfixed_const(wm->num_heads); a.full = dfixed_div(a, b); + tmp = div_u64((u64) dmif_size * (u64) wm->disp_clk, mc_latency + 512); + tmp = min(dfixed_trunc(a), tmp); - b.full = dfixed_const(mc_latency + 512); - c.full = dfixed_const(wm->disp_clk); - b.full = dfixed_div(b, c); - - c.full = dfixed_const(dmif_size); - b.full = dfixed_div(c, b); - - tmp = min(dfixed_trunc(a), dfixed_trunc(b)); - - b.full = dfixed_const(1000); - c.full = dfixed_const(wm->disp_clk); - b.full = dfixed_div(c, b); - c.full = dfixed_const(wm->bytes_per_pixel); - b.full = dfixed_mul(b, c); - - lb_fill_bw = min(tmp, dfixed_trunc(b)); + lb_fill_bw = min(tmp, wm->disp_clk * wm->bytes_per_pixel / 1000); a.full = dfixed_const(max_src_lines_per_dst_line * wm->src_width * wm->bytes_per_pixel); b.full = dfixed_const(1000); diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c index b420051e9da4..e10c82f7a37a 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c @@ -974,23 +974,10 @@ static u32 dce_v8_0_latency_watermark(struct dce8_wm_params *wm) a.full = dfixed_const(available_bandwidth); b.full = dfixed_const(wm->num_heads); a.full = dfixed_div(a, b); + tmp = div_u64((u64) dmif_size * (u64) wm->disp_clk, mc_latency + 512); + tmp = min(dfixed_trunc(a), tmp); - b.full = dfixed_const(mc_latency + 512); - c.full = dfixed_const(wm->disp_clk); - b.full = dfixed_div(b, c); - - c.full = dfixed_const(dmif_size); - b.full = dfixed_div(c, b); - - tmp = min(dfixed_trunc(a), dfixed_trunc(b)); - - b.full = dfixed_const(1000); - c.full = dfixed_const(wm->disp_clk); - b.full = dfixed_div(c, b); - c.full = dfixed_const(wm->bytes_per_pixel); - b.full = dfixed_mul(b, c); - - lb_fill_bw = min(tmp, dfixed_trunc(b)); + lb_fill_bw = min(tmp, wm->disp_clk * wm->bytes_per_pixel / 1000); a.full = dfixed_const(max_src_lines_per_dst_line * wm->src_width * wm->bytes_per_pixel); b.full = dfixed_const(1000); -- cgit v1.2.3